我国电力体制改革不断推进，现货市场建设进程日益加快，高效的电力市场出清技术是电力现货市场稳健运行的关键支撑。我国典型省级电网的日前市场出清是拥有数万个0-1变量和数十万条约束的大规模机组组合优化问题，面临“组合爆炸”难题，若直接调用商业数学优化软件求解，计算时间过长，难以满足日前市场决策的时效性要求，已经成为我国现货市场建设的难点和痛点。为此，本文以机组组合问题为对象，以算法突破为核心，围绕如何提升寻优有效性开展研究，从搜索有效性和约束条件有效性两个方面，分别提出了提升搜索有效性的0-1混合整数规划求解算法和基于不敏感约束辨识的机组组合求解算法，并将其应用于电力日前市场出清求解，极大地提升了计算效率。本文首先提出了提升搜索有效性的0-1混合整数规划求解算法。在搜索范围方面，提出了松弛邻域搜索算法，定义了松弛距离以度量任一解与松弛最优解的距离，实现在松弛最优解的邻域进行搜索。在搜索方向方面，提出了对称诱导函数算法，诱导整数变量趋整，实现了双向、对称、有效诱导。该算法能够在不改变最优性的前提下，通过有效搜索范围和搜索方向的选择，加快寻优进程，削减分支规模，避免了空间和路径遍历下组合优化的低效，大幅提高0-1混合整数规划问题的求解效率。算例表明，机组组合问题的计算时间减少了70%以上。本文提出了基于不敏感约束辨识的机组组合求解算法。当前的不起作用约束辨识方法大多仅能削减与可行域无关的冗余约束，已有研究表明，决定最优解的仅是最优解附近的敏感约束，如果能够削减与可行域相关、但距离最优解较远的不敏感约束，则将显著减小问题规模，显著提升求解效率，而现有方法仅能通过反复迭代实现。基于此，本文所提方法同时削减冗余约束和不敏感约束，仅保留最优解附近的敏感约束，无需反复迭代即可显著提升约束条件的有效性，算例表明，所提算法削减了90%以上的约束，大幅精简了模型规模，求解效率提升50%以上。本文提出了约束条件弹性化的机组组合高效求解算法。构建了约束条件弹性化、水火联合优化、火电机组深度调峰与可再生能源削减权衡互置的机组组合模式和模型；针对随之而来的模型规模剧增，提出了提升寻优有效性的机组组合高效求解算法。算例表明，应用该算法后，机组组合的求解效率提升90%以上。本文的研究工作已成功应用于我国广东、甘肃等地电力市场，为我国电力市场建设和出清效率的提高提供了理论与关键技术支撑。

With the development of China's power system reform and the acceleration of the spot market construction process, efficient power market clearing technology is the key support for the steady operation of the spot market. Day-ahead clearing of a typical provincial power system is a unit commitment problem with several ten thousand binary variables and several hundrend thousand constraints, which will be faced with combinatorial explosion problem. If it is solved directly by commercial optimization software, the computation time will be too long to meet the timeliness requirements of day ahead market optimization decision, which has become the difficulty and bottleneck of spot market construction in China. Therefore, this thesis takes unit commitment problem as the object, algorithm breakthrough as the core, and studies on how to improve the optimization effectiveness. A combinatorial optimization method for improving the optimization effectiveness is proposed. In the aspect of searching effectiveness, a 0-1 mixed-integer programming algorithm for improving the searching effectiveness is proposed. In the aspect of constraint effectiveness, a unit commitment algorithm for identifying the non-binding constraints is proposed. These algorithms are applied to day-ahead market unit commitment problems, which significantly improve the computational efficiency.A 0-1 mixed-integer programming algorithm for improving the searching effectiveness is first proposed in this thesis. In the aspect of searching space, a relaxation-based neighborhood search algorithm is proposed to explore the neighborhood of the linear program (LP) relaxation optimal solution. A new distance function, termed relaxation distance (RD), is proposed to measure the distance between the current solution and the LP relaxation optimal solution. In the aspect of searching direction, a symmetrical relaxation inducement algorithm is proposed to induce binary variables towards the tendency of the relaxed solution. The binary variables are symmetrically, bi-directionally and effectively induced. By selecting the effective searching space and searching direction, the proposed algorithm can accelerate the searching process, reduce the size of branch-and-bound tree, avoid the inefficiency of traversing the space and path and thus significantly improve the computational efficiency of 0-1 mixed-integer programming without changing the optimality of the solution. Case studies show that the unit commitment problem can reduce the computation time by at least 70%.A unit commitment algorithm for identifying the non-binding constraints is proposed. Most of the current inactive constraints identification methods can only reduce redundant constraints, which do not contribute to the feasible region. However, it is found that only sensitive constraints near the optimal solution determine the optimal solution. If the non-binding constraint, which contribute to the feasible solution but are far from the optimal solution, can be reduced, the computational efficiency can be improved markedly. However, the reduction of non-binding constraints can only be realized by iterations in current methods. Therefore, both redundant constraints and non-binding constraints are reduced and only sensitive constraints near the optimal solution are remained in this thesis. The constraint effectiveness can be improved significantly without iterations. Case studies show that the proposed algorithm can reduce at least 90% of transmission constraints, significantly reduce the scale of the unit commitment problem and thus markedly improve the computational effiency of unit commitment problem by at least 50%. An efficient constraint relaxation-based unit commitment algorithm is proposed. The mode and model of constraint relaxation-based unit commitment is proposed. Joint optimization of hydropower and thermal power, and the trade-off of the thermal unit deep ramp and the renewable energy curtailment, are realized. In response to the coming dramatic increase in model size, an efficient constraint relaxation-based unit commitment algorithm for improving the effectiveness of searching process is proposed. Case study shows that the proposed algorithm can improve the computational effiency by at least 90%.The research work of this thesis has been successfully applied to the electricity market of Guangdong province, Gansu province and other places in China and can provide theoretical and key technology support for the construction of electricity market and the improvement of clearing efficiency.